Particle Swarm Optimised polynomial neural network for classification: a multi-objective view

Classification using a Polynomial Neural Network (PNN) can be considered as a multi-objective problem rather than as a single objective one. Measures like predictive accuracy and architectural complexity used for evaluating PNN based classification can be thought of as two different conflicting objectives. Using these two metrics as the objectives of classification problem, this paper uses a Pareto based Particle Swarm Optimisation (PPSO) technique to find out a set of non-dominated solutions with less complex architecture and high predictive accuracy. The proposed method is used to train PNN through simultaneous optimisation of topological structure and weights. An extensive experimental study has been carried out to illustrate the importance and effectiveness of the proposed method.